MONETARY DEVELOPMENTS AND EXPANSIONARY FISCAL
*
CONSOLIDATIONS: EVIDENCE FROM THE EMU
António Afonso $, Luís Martins #
August 2014
Abstract
This paper provides new insights about the existence of expansionary fiscal consolidations in
the Economic and Monetary Union, using annual panel data for 14 European Union countries
over the period 1970-2012. Different measures for assessing fiscal consolidations based on
the changes in the cyclically adjusted primary balance were calculated. A similar ad-hoc
approach was used to compute monetary expansions in order to include them in the
assessment of non-Keynesian effects for different budgetary components. Panel Fixed Effects
estimations for private consumption show that, in some cases, when fiscal consolidations are
coupled with monetary expansions, the traditional Keynesian signals are reversed for general
government final consumption expenditure, social transfers and taxes. Keynesian effects
prevail when fiscal consolidations are not matched by a monetary easing. Panel probit
estimations suggest that longer and expenditure based consolidations contribute positively for
its success, while the opposite holds for the tax based ones.
Keywords: fiscal consolidation, monetary expansion non-Keynesian effects, panel data, probit
JEL: C23, E21, E5, E62, H5, H62
*
We thanks participants at the UECE 3rd Conference on Economic and Financial Adjustments for useful
comments and suggestions. The opinions expressed herein are those of the authors and do not necessarily reflect
those of the ECB or the Eurosystem.
$
ISEG/ULisboa –Universidade de Lisboa, Department of Economics; UECE – Research Unit on Complexity
and Economics, R. Miguel Lupi 20, 1249-078 Lisbon, Portugal, email: [email protected]. UECE is supported
by the Fundacão para a Ciência e a Tecnologia (Portuguese Foundation for Science and Technology) through the
PEst-OE/EGE/UI0436/2011 project, European Central Bank, Directorate General Economics, Kaiserstraße 29,
D-60311 Frankfurt am Main, Germany.
#
ISEG/ULisboa –Universidade de Lisboa, R. Miguel Lupi 20, 1249-078 Lisbon, Portugal. email: Luís Martins
[email protected].
1
1. Introduction
Keynesian theory gives us some insights about the expected effect of government
budgetary components’ changes on income. It postulates that an increase in government
spending should stimulate the economy, via the multiplier mechanism, thus increasing
disposable income and private consumption. Under this reasoning, an increase in taxation
should lead to a decrease in private consumption.
Nevertheless, since the early 90’s with the case studies of Denmark and Ireland 1, some
literature has been discussing the possible non-Keynesian effects of fiscal policy, namely
during fiscal consolidation periods.
The theoretical underpinnings stemmed from the German Council of Economic
experts in their 1981 and 1982 reports and are referred to as the “expectational view of fiscal
policy”.2 Arguably, the standard Keynesian relationship between private consumption and
government budgetary components may be reversed under certain circumstances. A
deterioration of the fiscal position (resulting in a budget deficit) today may lead to an increase
in taxation in the future in order to fulfil the government budget constraint, therefore reducing
the agents’ permanent income. If such expectations are accommodated by the individuals, this
could lead to a decrease in private consumption today. The reverse reasoning holds for a fiscal
consolidation, meaning that an improvement in the fiscal position may lead to an increase in
private consumption today. Some empirical research presents evidence that supports this
view.3
The expectational view of fiscal policy relies on the assumption of Ricardian
households, which smooth consumption and don’t have liquidity constraints. This motivates a
thorough assessment of the monetary developments while studying expansionary fiscal
consolidations. Moreover, according to the Keynesian view, under the IS-LM framework, a
fiscal consolidation may lead to an increase in private consumption if accompanied by a
strong enough monetary expansion that offsets the detrimental effects of fiscal policy
developments on disposable income and private consumption.
Arguably, while neglecting the monetary policy stance, one could find himself in a
situation described by Ardagna (2004): “In this case, the coefficients of fiscal policy variables
can be biased, capturing the effect of monetary rather than fiscal policy”.
1
See Giavazzi & Pagano (1990).
See Hellwig & Newmann (1987).
3
See for instance, Giavazzi & Pagano (1990), Perroti (1999), Ardagna (2004), Afonso (2006, 2010) and Alesina
& Ardagna (2013).
2
2
The importance of this issue within the Economic and Monetary Union (EMU) context
is rather obvious, since the expectational view of fiscal policy was to some extent reflected in
the fiscal convergence criteria of the Maastricht Treaty. Additionally, the monetary policy
stance is outside the national governments’ influence.
This paper contributes to the existing literature by providing some new insights about
the importance of the monetary stance for the relationship between fiscal developments and
private consumption during fiscal consolidation periods. It does so by expanding notably
Afonso’s (2006, 2010) and Afonso & Jalles’s (2014) core specification in order to
accommodate the monetary policy developments. We conduct an assessment of the fiscal
episodes using the same criteria. However, and in addition, we also identify monetary
episodes for 14 European Union countries from 1970 to 2012 and study their relationship with
fiscal developments.
The paper is organized as follows.
Section two presents the main theoretical underpinnings that motivated the study of
expansionary fiscal consolidations, followed by some literature review on this topic. Section
three presents an identification of the fiscal and monetary episodes and their respective
relationship. In section four we conduct the empirical analysis of expansionary fiscal
consolidations for the EMU resorting to panel estimations, accommodating the developments
of monetary policy, followed by a discussion of the results.
We also assess the factors that may impinge on the success of the fiscal
consolidations, namely expenditure based versus revenue based consolidations, using the
fiscal and monetary episodes identified in the earlier sections. We do so by relying on a probit
estimation based on Afonso & Jalles (2012) and a success definition proposed by Alesina &
Ardagna (2013).
Lastly, section five concludes with some final remarks and point out some possible
subjects for future research on this topic.
2. Literature survey
Hellwig & Neumann (1987) were pioneers regarding the assessment of the
expansionary fiscal consolidation hypothesis. They argue that fiscal consolidation in Germany
in the 1980’s under Chancellor Kohl had such a positive impact on private sector confidence
that demand actually increased. Supposedly, fiscal consolidation by the Federal Government
and monetary tightness by the Bundesbank led to continued output growth and low inflation.
Also lower deficits stimulated private investment in the long run due to reduced cost of
3
financing. Nevertheless, unemployment remained high, which authors attribute to labour
market rigidity.
Giavazzi & Pagano (1990) test this hypothesis for Denmark and Ireland for mid and
late 1980's, respectively. For Denmark they report that the thriving in consumption
experienced in 1983-1986 cannot be explained by the decline in interest rates alone and that
such is related to fiscal consolidation through the increase in revenue from income taxation
and decrease in public investment. Regarding the Irish case, the fast consumption growth in
the second stabilization was due to the government focus on decreasing spending, instead of
increasing taxation and also due to the liberalization of the credit markets. As a whole, in
these cases the expansionary fiscal consolidation is linked to an adjustment on the public
spending side, rather than on revenues, although in Denmark the adjustment was through
investment spending and in Ireland it was through current spending.
Alesina & Ardagna (1998) investigate the expansionary fiscal consolidation possibility
recurring to an analysis of OECD countries from 1960 to 1994. According to the general
Council of Economic Experts’ expectational view of the fiscal policy addressed in Hellwig
and Neumann (1987), fiscal adjustments occurring when the debt level is high or growing
rapidly should be expansionary, whereas others should not. Nevertheless the authors don’t
find evidence that confirm this view. On the other hand, they found strong evidence of the
effect of the composition of the adjustment in the outcome of the fiscal consolidation: all of
the non-expansionary adjustments were tax based and all of the expansionary ones were based
on expenditure cuts. Expenditure adjustments that were accompanied by wage moderation
and by nominal exchange rate devaluation turned out to be expansionary.
Perroti (1999) addresses the same issue for nineteen OECD countries from 1965 to
1994 and, according to his findings, substantial deficit cuts can lead to booms in private
consumption. The likelihood of an expansionary fiscal consolidation increases on times of
“fiscal stress”, which the author defined as periods of high debt-to-GDP ratio or following
periods of exceptionally high debt accumulation rates. His findings differ for other periods
since in “normal” times the Keynesian effects of a fiscal consolidation (either through
spending cuts or tax increases) on private consumption dominate.
Giavazzi et al. (2000) address the issue of expansionary fiscal consolidation in OECD
countries from 1973 to 1996 and in developing countries from 1960 to 1995. In OECD
countries evidence of non-Keynesian response by the private sector is more likely to be found
when the fiscal impulses are large and persistent. This means that only those can signal a
regime change, thus affecting private sector expectations. Also non-Keynesian effects leading
4
to an expansionary fiscal consolidation are stronger for changes in net taxes rather than in
public expenditure. In developing countries, non-Keynesian effects occur not only during
periods of fiscal contractions but also during fiscal expansions and when countries are piling
up debt rapidly, regardless of its level.
Using panel data from OECD countries from 1970 to 2002, Ardagna (2004)
investigates the effect of fiscal consolidations on debt-to-GDP ratio and GDP growth.
Regarding the debt-to-GDP ratio, the success of the fiscal consolidation depends more on the
size of the adjustment rather than its composition. On the other hand, the likelihood of a fiscal
consolidation being expansionary increases when it is based on public spending cuts rather
than on increased taxation. Concerning the role of the monetary policy, there was evidence
that either successful (leading to decrease in debt-to-GDP ratio) or expansionary (leading to
increase in GDP growth) consolidations don’t need to be met by expansionary monetary
policies nor exchange rate devaluations.
Giudice et al. (2004) address the matter of non-Keynesian effects in fourteen
European Union countries in an ex-post and ex-ante analysis. Ex-post analysis consisted on
studying the period from 1970 to 2002 and see if fiscal consolidation episodes were followed
by an increase in GDP growth. Results show that this occurred in about half the cases. The
ex-ante analysis carried out was based on simulations by the European Commission QUEST
model and suggested that short-term non-Keynesian effects can occur if consolidation is
mainly on the spending side. The latter is also true in the ex-post case, which is in line with
most of empirical studies.
Afonso (2006, 2010) conducted a panel analysis for 15 EU countries from 1970 to
2005 having found some evidence of non-Keynesian effects in private consumption for some
government spending items, namely final consumption and social transfers. Results show that
a decrease in government consumption leads to an increase in private consumption in the long
run and the magnitude of this effect is higher when a fiscal consolidation episode occurs.
Devries et al. (2011) construct a database for fiscal consolidation measures taken by
17 OECD countries from 1978 to 2009, based on the premise that computing fiscal
consolidations from the changes of the cyclically adjusted primary balance may be
problematic. Arguably, such approach may be biased in the sense that it may capture changes
that are not related to policy actions due to its inability to remove sharp fluctuations in
economic activity. Therefore they identify fiscal consolidations through an historical
5
approach based on policy documents. This database has been widely used in subsequent
literature concerning expansionary fiscal consolidations.4
Afonso & Jalles (2012) analyse a panel of OECD countries from 1970 to 2010 to
assess if the composition and duration of fiscal consolidations matter for their success.
Consolidation episodes lead to a decrease in the debt ratios only if accompanied by strong
economic growth and increased output gap. Increased duration contributes to the success of
the fiscal consolidation episode. Fiscal consolidation success depends on the composition of
the adjustment: consolidations based mainly on tax increases contribute negatively for its
success.
Alesina & Ardagna (2013) use Devries et al. (2011) policy action based approach to
identify the fiscal episodes for 21 OECD countries from 1970 to 2010. They conclude that
expenditure based adjustments are more likely to be successful and expansionary. Monetary
policy is not significant to explain the differences between expenditure based and tax based
adjustments.
To sum up, most of the research seems to support or at least not reject the idea of
expansionary fiscal episodes.5 Also, some findings6 suggest that expansionary and successful
fiscal episodes are more likely when there is a consolidation on the spending side.
Moreover, some of the literature, such as Perroti (1999) and Giavazzi et al. (2000)
propose that non-Keynesian effects are more likely or only occur during periods of high debtto-GDP ratio or when debt is piling up quickly.
3. Identification of fiscal and monetary episodes in the EMU
3.1. Fiscal Episodes
Most of the empirical literature relies on the change in the cyclically adjusted primary
balance (CAPB) as a percentage of GDP as measure of the governments’ structural budget
balance. It extracts the elements of the primary balance that are due to the business cycle from
the total balance in order to have an indicator that has been corrected from the effects of
changes in economic activity, thus reflecting the discretionary part of the fiscal policy. Table
XI in the appendix shows some descriptive statistics of this indicator.
One can assess the existence of fiscal episodes – either contractions or expansions – by
studying the behaviour of this indicator over time. In Giavazzi & Pagano (1996) there is a
4
See for instance Afonso & Jalles (2012) and Alesina & Ardagna (2013).
As seen in Giavazzi & Pagano (1990), Alesina & Ardagna (1998), Perroti (1999), Giavazzi et al. (2000), van
Aarle & Garretsen (2003), Ardagna (2004), Giudice et al. (2004) and Afonso (2006, 2010).
6
Giavazzi & Pagano (1990), Alesina & Ardagna (1998), Afonso (2006, 2010) and Alesina & Ardagana (2013).
5
6
fiscal episode when the cumulative change in the cyclically adjusted primary balance is at
least 5, 4, or 3 percentage points of GDP in 4, 3 or 2 years respectively or 3 percentage points
in one year. Alesina & Ardagna (1998) identify the periods of occurrence of fiscal episodes
by looking for the periods when the change in the cyclically adjusted primary balance was
greater than 2 percentage points in one year or at least 1.5 percentage points of GDP on
average in the last two years. Afonso’s (2006, 2010) assessment of fiscal episodes relies on a
different method: there is a fiscal episode when the change in the cyclically adjusted primary
balance is greater than 1.5 times the panel standard deviation of this indicator or when the
average absolute change in the last two years is greater than the standard deviation of the full
panel. Table I shows the fiscal expansions and contractions according to the different criteria.7
The measures used by Giavazzi & Pagano (1996), Alesina & Ardagna (1998) and
Afonso (2006, 2010), were labelled respectively as FE 1 , FE 2 and FE 3 . Overall there is a
considerable overlapping of episodes according to the different criteria: there is a coincidence
of 82 and 63 percent between fiscal episodes 1 and 2 and 1 and 3, respectively and 82 percent
between criteria 2 and 3 (see Table I). The highest number of episodes is given by the criteria
used by Alesina & Ardagna (1998), although the methodology followed by Giavazzi &
Pagano (1996) leads to higher duration in both expansions and contractions.
All the criteria capture the cases studied by Giavazzi and Pagano (1990), as there is an
identification of fiscal contractions in Denmark in 1983-86 and in Ireland in 1988. Also, there
is a clear identification of fiscal expansions in 2009 across the EMU countries, following the
European Commission policy recommendations after the 2007-08 financial crisis.
Furthermore, the different methodologies also identify the consolidation efforts made by the
countries under financial assistance in 2011-2012, namely Ireland, Greece and Portugal.
7
We used a slightly lower threshold for the Afonso’s (2006, 2010) methodology, due to the increase in the
standard deviation of the panel sample from 1.57 to 2.00. We used 1, instead of 1.5 times the standard deviation.
7
Table I - Identification of the fiscal episodes according to the different criteria
(1970-2012)
Country
Austria
Belgium
Denmark
Finland
France
Germany
Greece
Ireland
FE 1
Expansions Contractions
04
97
81, 05, 09
82-87
75-76,
83-87
90-91
79-80, 83,
76-77,
91-93, 10
97-98, 00-01
75, 91, 95,
01-02
04, 08-09
01-02,
07-11
Italy
FE 2
Expansions Contractions
04
84, 97, 01, 05
81, 05, 09
82-85, 06
75, 82,90
83-86
78,87, 91,
09
83, 92-94, 12
09
75, 90-91,
95, 01, 10
89, 95,
08-09
95, 01,
07-10
81, 01
96-99, 12
92-94, 96,
10-12
88, 11-12
76-77, 81,
88, 96-97,
00-01
96-97, 00,
11-12
91-92, 94,
10-12
88, 11-12
82-83,
92-93, 12
91, 93, 96
FE 3
Expansions
Contractions
04
84, 97, 01, 05
81, 05, 09
82, 84-85, 06
75, 82, 90
83-86
78, 87, 91-92,
10
09
75, 90-91, 95,
01-02, 10
89, 95, 08-09
95, 01-02,
07-10
81, 01
01, 09
76, 88, 96, 00
96-97, 00, 11
91-92, 94, 1012
88, 11-12
82-83, 92-93,
12
91, 93, 96
Netherlands
02, 09-10
91, 93
01, 09
Portugal
83-84, 11-12
78-79, 85,
93, 05, 09
08-09
83-84, 86,
88, 92, 11-12
78, 85, 93,
05, 09-10
08-09
83, 86, 88, 92,
11-12
Spain
78-80, 94,
09-10
08-11
Sweden
United
Kingdom
02-03
91-93,
01-04, 09
96-99
97-00, 11-12
02
90, 92,
01-02, 09
96-97
97-98, 00,
11-12
02
90, 92-93,
01-03, 09
96-97
00, 11
# Years
with
episodes
Average
duration of
episodes
(years)
53
55
62
57
51
46
1.89
2.39
1.63
1.63
1.34
1.35
Source: author’s computations. Notes: FE 1 - Measure based on Giavazzi & Pagano (1996); FE 2 - Measure
based on Alesina & Ardagna (1998); FE 3 - Measure based on Afonso (2006, 2010).
Recent studies such as Afonso & Jalles (2012) and Alesina & Ardagna (2013) also
include a criterion for identifying fiscal consolidations referred to as IMF’s “Action Based
Approach” computed by Devries et al. (2011). It identifies fiscal consolidations based not on
the changes in CAPB, but on an historical approach through the analysis of policy documents.
Arguably the CAPB based fiscal consolidations may be biased in the sense that they may
capture changes that are not related to policy actions due to its inability to remove sharp
fluctuations in economic activity. Unfortunately the database is still under update, so we
would have to discard the most recent years (2010-2012) in order to accommodate that
approach. Therefore we will not include it at this point, but we intent to do so in future
research.
8
3.2.Monetary episodes
One of the main points in this paper is the study of the coupling of fiscal and monetary
policy in order to assess if monetary expansions have an impact on the relationship between
government budgetary components and private consumption during fiscal consolidation
episodes. Therefore, it is crucial to establish a clear identification of the monetary episodes in
the EMU countries. We chose three indicators that could be used as a measure of the
monetary stance for the different countries, namely the real short term money market interest
rate, and the nominal and real effective exchange rates.
The change in the real short term interest rate is a widely used measure of the
monetary policy easing or tightening8 as it accounts not only for the money market rates but
also for the price developments. Therefore a negative variation in this indicator signals a real
monetary easing, rather than a nominal one.9
Both the nominal and the real effective exchange rate assess the currency value in a
country vis-à-vis a weighted average of other selected countries’ currencies, being commonly
used to assess the countries’ competitiveness. The nominal effective exchange rate has been
used by Ardagna (2004) as an indicator of the monetary stance. A negative change in this
indicator corresponds to currency depreciation and therefore a monetary expansion. We also
included the real effective exchange rate with the purpose of accounting for possible
differences in monetary episodes identification due to price developments, which links to the
arguments presented on the interest rates case.
In order to define the monetary episodes we relied on a similar strategy as Afonso
(2006, 2010) and identified an episode when the absolute change in one year or the average
change in two years in the different indicators was greater than 1.5 times or 1 time the panel
standard deviation respectively:
1, if M tl  1,5 l


M tl  M tl1
MEtl  1, if
l
2

0, otherwise

8
l  1, 2,3 .
(1)
See for instance Afonso & Sousa (2011).
Since the nominal short-term interest rates are very similar in the EMU countries from 1999 onwards we
cannot include them in my estimations due to near singular matrix issues and so they were excluded from this
analysis.
9
9
MEtl denotes a monetary episode in period t according to criteria l; M tl corresponds
to the change of the indicator l in period t. For the real short term interest rate we have an
absolute change while in the nominal and real effective exchange rates we used the
percentage change of the respective indexes.
stands for the panel standard deviation of the
relevant indicator.
Table II shows the monetary episodes identified according to the different indicators.
ME1 , ME 2 and ME 3 correspond to the use of the methodology across the changes in the real
short term interest rate, and the percent changes in the real and nominal effective exchange
rate, respectively.
One of the main highlights is that there are considerably more monetary episodes than
fiscal ones. The duration of the monetary episodes also changes significantly across the
different criteria. If we look at the monetary episodes based on the change in the real short
term interest rate ( ME1 ), it is possible to see that the expansions and contractions last 1.5 and
1.8 years on average respectively. If we consider the changes in the nominal effective
exchange rates the duration of the expansions more than doubles and in the case of the
contractions it also increases significantly.
Moreover, while in the fiscal episodes case there is a significant overlapping across the
different criteria, in here it is much lower, with the matching being only 38, 51 and 63 percent
between ME1 and ME 2 , ME1 and ME 3 and ME 2 and ME 3 , respectively. The splitting
between expansions and contractions is fairly even, with the exception of ME 3 , which reports
considerably more contractions than expansions. Also, we can see that there are episodes
labelled as expansions in ME1 that show up as contractions in ME 2 and ME 3 , which further
motivates the inclusion and analysis of all the different criteria.10
The descriptive statistics of the indicators used to identify both fiscal and monetary
episodes can be consulted in table XI in the Appendix.
10
For instance in Austria we have a monetary expansion in 1983 expansion according to ME1 , but it shows up as
a contraction in ME 3 .
10
Table II – Identification of the monetary episodes according to the different criteria
(1970-2012)
Country
Austria
Belgium
Denmark
ME 1
Expansions Contractions
72, 83, 94,
77, 80-81,
09-10
89-90
72,75,
76-77, 79-81,
82-83,
90-91
93-94, 10
73, 81,
76-78, 90-91,
94-97, 10
93, 07, 11
ME 2
Expansions Contractions
97-98, 00
77, 80, 87,
93, 95, 04
81-83,
77, 79, 86-87,
97-98, 00
95, 03-04
80-82, 00
79, 86-87,
03-04, 09
74-76, 80-82,
85, 89-90,
95-96, 03-04
86-87, 03-04
86, 89, 92-94,
98
72, 78-79,
92-94, 97,
00, 11
82-84,
97-98, 0001
81-82, 85,
89, 97-98,
00-01, 11
83-86, 0001
75-76, 81,
88-89,
92-94,
98-99, 1012
73-74, 94,
99, 09
74, 77-79,
83-85, 90-91,
07-09
88-89, 9394, 99-00,
10-12
76, 81-85, 92
93-95, 00
71-72,
94-95, 10
73-75, 80,
83, 88, 9495, 98, 10
84-86, 88,
95, 99
86-87, 9394
73-74, 78-80,
90, 07
76-79, 81-82,
85, 87, 90-91,
08
78-81, 83,
87-88, 07-08
85, 92-93
81, 84-85,
89, 97, 00
77-80, 8384
United
Kingdom
74-75, 88,
02, 09-10
73, 76-77,
81-82, 90, 98
# Years
with
episodes
Average
duration of
episodes
(years)
96
1.55
Finland
France
Germany
Greece
Ireland
Italy
Netherlands
Portugal
Spain
Sweden
71-74, 88,
93-95, 98,
12
72, 75-76,
94, 97
75, 82-83,
86, 93, 02,
09-10
82, 90, 9596, 00-03
75-76, 80,
83-84, 89-92
74, 77, 81, 90
73, 80-81, 90
79, 87, 93-95,
03-04
82, 88, 90-91,
95-96, 03-04,
08
79-80, 82-83,
86-87, 02-04,
07-08
ME 3
Expansions
Contractions
73-80, 83, 8688, 93, 95
81-83, 97
77-78,
86-87, 91, 95,
03-04
80-82, 00
73-74, 76, 8687,
90-91, 93, 95,
03-04
72-73,
81, 89-90, 9478-79, 92-93,
96, 03-04
97, 00
77-78,
73, 75-76, 8681-84, 00
87, 90, 93-96,
03-04
97, 00
72-80, 83-84,
86-88, 93-96,
03-04
72-95
03-04
73-77,
81-82, 84, 992000
86, 90-91, 0304, 08
83-84, 86-87,
90-91, 96-97,
03-04
77, 79, 87,
95, 02-04
81-82, 89-93,
02-04
73-85,
93-95, 00
87, 96-97, 0304
97
74-78, 83, 8688, 93-95
85-91, 02-03,
08
79-80, 85,
89-91, 96,
03-04, 10-12
80-81, 88-89,
91, 97-99, 05,
07, 11-12
76-78,
81-84, 93-94
78-79, 82-84,
93-94, 01-02, 09
74, 79,
89-91, 03-04
76, 96-97, 0304, 10-12
73-77,
83-84, 87, 9394, 08-10
79-81, 88, 9799
92
82-84, 9394
78, 82-84,
93-94,
98-02, 09
83-84,
86-87,
93-94, 0810
95
124
124
122
1.80
1.98
1.85
3.26
2.22
76-89, 94
Source: author’s computations. Notes: ME 1 - Measure based on the changes in the real short term interest rate;
ME 2 - Measure based on changes in the real effective exchange rate; ME 3 - Measure based on the changes in
the nominal effective exchange rate.
11
4. Empirical assessment
4.1. Data description
The data consists on annual frequency time series ranging from 1970 to 2012 for
private consumption, GDP, general government final consumption, social transfers, taxes,
cyclically adjusted primary balance, general government debt, revenue and expenditure, taken
from the AMECO database.11 We used 11 countries who belong to the EMU,12 namely
Austria, Belgium, Germany, Finland, France, Greece, Ireland, Italy, Netherlands, Portugal
and Spain and also Denmark, Sweden and United Kingdom, which are not in the EMU, but
are geographically and politically linked to the remaining. This means that we can have a
maximum of 602 observations per variable, throughout the entire panel.
Tables XI-XIII in the Appendix show the descriptive statistics and unit root tests for
the series used in the estimations.
The unit root tests in table XII in the Appendix show that most series are stationary.
For the ones that are not, since we have already computed significant changes on the original
series, such that what we have are the logarithms of the real per capita values, it makes sense
to include all the series in levels. Otherwise we would risk losing some of the intuition behind
the variable relationship, thus making the model more difficult to interpret.13
4.2. Modelling expansionary fiscal consolidations
The strategy for accessing the potential differences between fiscal expansions and
fiscal contractions is based on Afonso (2006, 2010). It consists on estimating the variation of
private consumption, using budgetary variables and dummies for assessing fiscal and
monetary episodes. The core specification will be
Cit  ci  Cit 1  0Yit 1  1Yit   0Yitav1  1Yitav 
(1 FCEit 1   3FCEit  1TFit 1  3TFit   1TAX it 1   3TAX it )  FCitm 
(2)
( 2 FCEit 1   4 FCEit   2TFit 1   4 TFit   2TAX it 1   4 TAX it )  (1  FCitm )  it
11
For full description of the original series see table X in the Appendix.
Originally we also had Luxembourg, which was dropped due to the lack of information on monetary data.
13
My argument follows the explanation presented in Afonso (2006, 2010).
12
12
where i (i  1,..., N ) indicates the different countries, t (t  1,..., T ) stands for the period. We
also have: C – private consumption; Y – GDP; Y av – panel’s GDP average;14 FCE – general
government final consumption expenditure; TF – social transfers; TAX – taxes. All variables
displayed correspond to the natural logarithm of the real per capita values.15 FC m is a
dummy variable that identifies a fiscal consolidation episode, according to the three different
criteria mentioned in the previous section (m  1, 2,3) . Therefore, when FCitm is equal to one,
there is a fiscal consolidation in period t, for country i, according to the criteria m. ci is an
autonomous term which captures each country’s individual characteristics, being the source of
cross-country heterogeneity in a Fixed Effects model, which will be our estimation choice.
The disturbances it are assumed to be independent and identically distributed across
countries with zero mean and constant variance.
4.2.1. Core specification outputs
According to Greene (2012), we use the Fixed Effects (FE) estimation whenever we
are interested in analysing the impact of variables that change over time. It explores the
relationship between predictor and dependent variables within a country. The FE model
removes the effect of time-invariant characteristics from the predictor variables so we can
assess the independent variables net effect. An important assumption of the model is that time
invariant characteristics are country specific and should not be correlated with other
individual features. In other words, each country has unique attributes that are not the result of
random variation and that do not vary across time. The source of country heterogeneity is
given by the intercept ci in specification (1) with Fixed Effects allowing for correlation
between the latter and the repressors.16
We perform redundant FE likelihood ratio tests for all estimations, with the null
hypothesis being that there is no unobserved heterogeneity and so the model can be estimated
by pooled OLS. If we reject this hypothesis then fixed effects is more adequate than pooled
14
The original specification in Afonso (2006, 2010) used the OECD’s GDP instead of the panel average.
Nevertheless, since OECD only displays that series starting from 1995 we followed Afonso & Jalles (2011) and
used the panel average GDP.
15
For instance, in order to obtain the variable
Y we make the following calculations:
 GDP / DEF 
Y  ln 
 , where
N


GDP stands for the GDP at current prices, DEF and N correspond to the
GDP deflator and total population, respectively.
16
In the FE estimation the intercept also works as a substitute for non-specified variables, yielding consistent
estimates in the presence of correlation between the latter and the repressors, which favours the usage of this
model in comparison to pooled OLS.
13
OLS, since it allows for cross country heterogeneity by permitting each one to have its own
intercept value ( ci ).17
Table III presents the estimation results for specification (2) according to the different
criteria for identifying fiscal consolidation episodes. Both consumption and income are
statistically significant across the different specifications. The negative sign for consumption
in t-1 (  ) has obviously to do with the fact that we have the lagged consumption as an
independent variable, therefore increasing consumption in period t-1 decreases its difference
between t and t-1. The short-run elasticity of private consumption to income is similar across
specifications, ranging between 0.083 and 0.087.
There is a positive statistically significant relationship between the first difference of
general government final consumption expenditure ( FCEt ) and private consumption ( Ct )
when we have a fiscal consolidation ( FC m  1 ), across all of the estimations based on (2),
with coefficients between 0.193 and 0.237. Such relationship is in line with the traditional
Keynesian effects, indicating that consumers are not behaving in a Ricardian way, since they
do not seem to anticipate the need for increased taxation in the future due to an increase in
government spending today.
The previous relationship does not hold in the absence of a fiscal consolidation
episode. Moreover, there is some evidence of non-Keynesian effects in the absence of fiscal
consolidations ( FC m  0 ) if we look at the final consumption expenditure ( FCEt 1 ) and taxes
( TAX t 1 ) in column 3 and across the three different estimations, respectively. The negative
sign in the short-run elasticity of general government final consumption expenditure to private
consumption suggests Ricardian behaviour in the absence of fiscal consolidations. Similar
non-Keynesian reasoning holds for the relationship between taxes and consumption, meaning
that an increase in taxes today leads to increased spending as consumers anticipate that there
is no need for increased taxation in the future.
17
We report the redundant FE likelihood ratio for all estimations. In any case the no cross-country heterogeneity
assumption is always rejected, meaning that the FE estimator is more adequate than pooled OLS.
14
Table III – Fixed Effects estimation results for specification (2)
FC 1
-0.090***
(-3.62)
0.084***
(3.06)
0.809***
(11.41)
-0.026*
(-1.80)
-0.169**
(-2.38)
FC 2
-0.089***
(-3.61)
0.083***
(3.06)
0.812***
(11.51)
-0.025*
(-1.74)
-0.165**
(-2.31)
FC 3
-0.076***
(-3.05)
0.087***
(3.14)
0.835***
(12.24)
-0.033**
(-2.19)
-0.187***
(-2.62)

Ct 1
0
Yt 1
1
Yt
0
Yt av
1
1
Yt av
1
FCEt 1
0.005
(0.20)
0.010
(0.50)
-0.001
(-0.06)
3
FCEt
0.193*
(1.85)
0.214**
(2.05)
0.237**
(2.09)
1
TFt 1
0.003
(0.22)
0.002
(0.18)
-0.003
(-0,13)
-0.005
(-0.06)
0.016
(0.18)
0,063
(0,49)
 FC
m
3
TFt
1
TAX t 1
0.005
(0.23)
0.001
(0.07)
0,007
(0,33)
3
TAX t
0.044
(0.72)
0.033
(0.67)
0,027
(0,45)
2
FCEt 1
-0.013
(-0.98)
-0.015
(-1.16)
-0,027**
(-2,10
4
FCEt
0.048
(0.79)
0.041
(0.67)
0,007
(0,12)
2
TFt 1
0.002
(0.22)
0.002
(0.30)
0,000
(-0,07)
4
TFt
0.031
(0.97)
0.033
(1.05)
0,033
(1,07)
2
TAX t 1
0.024**
(2.15)
0.025**
(2.26)
0,029**
(2,43)
4
TAX t
0.033
(1.43)
454
0.73
t-stat.
3.09
p-val.
0.00
0.029
(1.21)
454
0.732
t-stat.
3.04
p-val.
0.00
0,033
(1,44)
440
0,742
t-stat.
3.04
p-val.
0.00
-1.74
0.08
1.54
0.12
1.35
0.18
-0.45
0.65
-0.20
0.84
0.09
0.93
 (1  FC m )
N
R2
Redundant FE likelihood
ratio
Null hypothesis
3   4  0
1   2  0
Notes: Used robust heteroskedastic-consistent standard errors. The t-statistics are in parentheses. *, ** and ***
denotes statistically significant at a 10, 5 and 1 percent level, respectively. FC 1 - Measure based on Giavazzi &
Pagano (1996); FC 2 - Measure based on Alesina & Ardagna (1998); FC 3 - Measure based on Afonso (2006,
2010).
15
However the Wald coefficient statistical tests suggest that there is no significant
difference between the presence and absence of fiscal consolidations concerning the short-run
effects of government final consumption expenditure and taxation on private consumption
(the null hypothesis: 3   4  0 and  1   2  0 are not rejected on the third and all
specifications, respectively).
Comparing to the literature that used similar methodology, as a whole our results
differ from Afonso (2006, 2010) and Afonso & Jalles (2012), since we find no evidence of
non-Keynesian effects considering general government final consumption expenditure or
taxes in the presence of fiscal consolidations ( FC m  1 ). However, our findings are similar
for periods of no fiscal consolidation ( FC m  0 ), since there is some evidence of nonKeynesian effects in this case for the mentioned budgetary variables.
4.2.2. Fiscal consolidations and monetary expansions.
The following specification is one of the main contributions of this paper, adding each
country’s monetary developments to specification (2). It will permit a breakdown of all the
possible combinations between fiscal contractions and monetary expansions, thus allowing
for the study of the possible differences between them.
Cit  ci  Cit 1  0Yit 1  1Yit   0Yitav1  1Yitav 
(10 FCEit 1   30 FCEit  10TFit 1  30 TFit   10TAX it 1   30 TAX it  50 M itl )  FCitm MX itl 
( 20 FCEit 1   40 FCEit   20TFit 1   40 TFit   20TAX it 1   40 TAX it  60 M ilt )  (1  FCitm ) MX itl 
(3)
(11 FCEit 1   31FCEit  11TFit 1  31TFit   11TAX it 1   31TAX it  51M itl )  FCitm (1  MX itl ) 
( 21 FCEit 1   41FCEit   21TFit 1   41TFit   21TAX it 1   41TAX it  61M itl )  (1  FCitm )(1  MX itl )  it
In addition to the repressors previously explained, MX itl denotes a monetary expansion in
period t (t  1,..., T ) for country i (i  1,..., N ) according to the criteria l (l  1, 2,3) . M l
corresponds to the relevant indicator used to calculate the monetary episodes on (1). We have
found some evidence of non-Keynesian effects during fiscal consolidations in 5 out of the 9
possible estimations.18 Tables IV and V show some of the most relevant estimation results.
18
Notice that since we have three different criteria for fiscal and monetary developments, the assessment of their
relationship within the current framework yields 9 possible estimation outputs. The other outputs are available in
tables XIV-XVII in the Appendix.
16
Table IV – Fixed Effects estimation for specification (3): 1st output

Ct 1
0
Yt 1
1
Yt
0
Yt av
1
1
Yt av
10
FCEt 1
 30
FCEt
10
TFt 1
 30
TFt
 10
TAX t 1
 30
TAX t
50
M tl
 FC m
 MX l
FC 1 , MX 3
FC 2 , MX 1
FC 3 , MX 1
-0.096***
(-3.80)
0.093***
(3.25)
0.803***
(10.93)
-0.019
(-1.22)
-0.181**
(-2.53)
0.050
(1.40)
-0.213***
(-3.58)
0.010
(0.69)
-0.130*
(-1.83)
-0.051**
(-2.06)
-0.132***
(-3.07)
0.096**
(2.08)
-0.097***
(-4.05)
0.099***
(3.64)
0.799***
(11.14)
-0.029**
(-2.06)
-0.155**
(-2.22)
0.200
(1.33)
-0.369*
(-1.72)
0.026
(0.84)
0.034
(0.12)
-0.206*
(-1.71)
0.484***
(4.55)
0.003
(0.49)
-0.097***
(-4.10)
0.102***
(3.82)
0.789***
(11.17)
-0.0294**
(-2.06)
-0.145**
(-2.08)
-0.840***
(-14.24)
-0.039
(-0.30)
1.293***
(21.38)
-11.42***
(-19.53)
-0.552***
(-9.29)
2.694***
(17.32)
-0.215***
(-20.12)
17
Table V – Fixed Effects estimation for specification (3): 1st output (cont.)
 20
FCEt 1
 40
FCEt
 20
TFt 1
 (1  FC m )
FC 1 , MX 3
-0.005
(-0.24)
0.270**
(2.51)
0.014
(1.21)
-0.043
(-0.91)
-0.007
(-0.47)
FC 2 , MX 1
-0.035**
(-2.15)
0.010
(0.09)
-0.018
(-1.57)
-0.028
(-0.58)
0.056***
(3.74)
FC 3 , MX 1
-0.038**
(-2.39)
0.010
(0.09)
-0.018
(-1.52)
-0.029
(-0.59)
0.055***
(3.67)
 40
TFt
 20
TAX t 1
 40
TAX t
-0.027
(-0.52)
-0.002
(-0.04)
-0.004
(-0.08)
60
M tl
0.025
(0.52)
-0.000
(-0.55)
-0.001
(-0.48)
11
FCEt 1
0.011
(0.42)
0.005
(0.22)
0.002
(0.07)
 31
FCEt
0.260**
(2.36)
0.310***
(3.57)
0.405***
(5.13)
11
TFt 1
 FC m
0.003
(0.35)
-0.005
(-0.37)
 31
TFt
 (1  MX l )
-0.007
(-0.40)
-0.093
(-1.01)
-0.025
(-0.31)
-0.063
(-0.59)
 11
TAX t 1
-0.006
(-0.28)
0.002
(0.09)
0.008
(0.40)
 31
TAX t
0.120
(1.58)
-0.017
(-0.36)
-0.051
(-0.85)
51
M tl
0.044
(0.84)
0.001
(1.31)
0.002
(1.78)
 21
FCEt 1
-0.021
(-1.42)
-0.015
(-0.98)
-0.017
(-1.17)
 41
FCEt
0.006
(0.10)
0.038
(0.55)
0.029
(0.42)
 21
TFt 1
0.003
(0.45)
0.005
(0.61)
0.005
(0.70)
 41
TFt
0.040
(1.09)
0.057
(1.64)
0.054
(1.60)
 21
TAX t 1
0.017
(1.36)
0.018*
(1.74)
0.016
(1.58)
 41
TAX t
0.029
(1.19)
0.046*
(1.83)
0.044*
(1.82)
61
M tl
0.036
(1.23)
454
0.759
t-stat.
p-val.
-0.001
(-0.93)
454
0.755
t-stat.
p-val.
-0.000
(-0.88)
454
0.763
t-stat.
p-val.
3.53
0.00
3.85
0.00
4.08
0.00
N
Redundant FE
likelihood ratio
 MX l
 (1  FC m )
 (1  MX l )
Notes: Used robust heteroskedastic-consistent standard errors. The t-statistics are in parentheses. *, ** and
*** denotes statistically significant at a 10, 5 and 1 percent level, respectively.
18
We can see that when the fiscal consolidations are matched by a monetary expansion
there is a negative and statistically significant short-term elasticity between the government
final consumption expenditure and private consumption ( 30  0 in the first and second
outputs and 10  0 in the third output). This doesn’t hold for fiscal consolidations that are not
accompanied by a monetary easing as  31 is positive and statistically significant and 11 is not
statistically significant across the respective outputs. The second and third estimation results
also show some evidence of non-Keynesian elasticity on taxes when there are both fiscal
contractions and monetary expansions (  30  0 ). Just like in the previous case, such effects
seem to disappear when we have fiscal consolidations without the respective monetary easing
as  31 is not statistically significant. The same pattern emerges again for social transfers on the
first and third outputs ( 30 is negative and statistically significant, but  31 is not statistically
significant). Table XIX in the appendix shows the summary of the robustness test computed
for estimations based on specification (3)
The Wald coefficient restriction tests, in table
XVIII in the Appendix, show that the difference between these coefficients is statistically
significant in all cases, except for social transfers in the first output ( 30  31  0 is not
rejected at a 10% level in this case).
A possible explanation relates to liquidity restrictions, which may prevent a Ricardian
behaviour, thus undermining the permanent income hypothesis. If households do have
liquidity constrains, a fiscal consolidation could signal indeed a future tax decrease and a
permanent income rise, which is perceived by the households, but does not materialize in a
current private consumption increase due to limitations in access to credit markets. Such is
summarized by Alesina & Ardagna (1998) as “the size of the increase in private consumption
[following government spending cuts] depends on the absence of liquidity-constrained
consumers”.
The IS-LM framework argument presented by Ardagna (2004) that the signs of the
coefficients may be biased in the sense that they are capturing the monetary stance is unlikely,
since we are controlling for these.
4.3.Measuring the success of fiscal consolidations
In this section we will investigate what are the factors that may contribute to the
success of fiscal consolidations. We computed dummy variables for successful fiscal
adjustments in two different ways based on the literature in order to assess if our findings are
19
robust across different criteria. The first measure ( SU t1 ) is based on Afonso & Jalles (2012)
who define a fiscal consolidation as being successful if the change in the cyclically adjusted
primary balance ( bt ) for two consecutive years is greater than the standard deviation (  ) of
the full panel sample:
1

1,
if
bt i  


.
SU t1  
i 0
0, otherwise

(4)
We have also included a measure computed by Alesina & Ardagna (2013) which is
based on the level of debt as a percentage of GDP. A fiscal consolidation is successful if the
debt-to-GDP ratio two years after the end of the fiscal adjustment ( Debtt  2 ) is lower than the
debt-to-GDP ratio in the last year of the adjustment ( Debtt ):
1, if Debtt  2  Debtt
SU t2  
0, otherwise
(5)
The identification of the leading policy option for the fiscal consolidation – either
expenditure or revenue based – is also assessed through dummy variables. Therefore, a fiscal
consolidation on period t is defined as being expenditure based ( EXPt ) if the change in the
total expenditure of the general government as a percentage of GDP in that period ( expt )
accounts for a proportion greater than  of the change in the cyclically adjusted primary
balance ( bt )
 expt


1, if
bt
EXPt  
.
0, otherwise

(6)
Following Afonso & Jalles (2012) we computed the composition of the adjustment for three
different thresholds, so that  assumes the values of 1/2, 2/3 and 3/4. A similar process was
conducted for the revenue based consolidations. Table VI shows the number of fiscal
consolidation episodes and their respective success rate (successes / total events), based on the
20
criteria defined in the earlier sections. The identification of the successful episodes follows
specifications (4) and (5). Table XX in the Appendix shows the successful fiscal episodes for
each country according to the different criteria.
Table VI – Fiscal consolidation events and success rates
FC1
FC 2
FC 3
SU 1
Total events Successes Success rate (%)
49
25
51%
56
32
57%
46
28
61%
Successes
29
19
17
SU 2
Success rate (%)
59%
34%
37%
Source: author’s computations. Notes: SU 1 - Measure of success based on (4); SU 2 Measure of success based on (5).
If we look at the successful events based on the change in CAPB for two consecutive
years, SU 1 , we can see that the success rates range from 51% to 61%, according to FC1 and
FC 3 , respectively. Although FC 2 reports the highest number of successful consolidations, the
success rate is slightly lower than in FC 3 (57% vs 61%), thus being penalized by the higher
number of consolidation events (56 vs 46). Nevertheless, the success rates are similar to the
ones found on Afonso & Jalles (2012).
In the case of SU 2 , we can see that there are significant differences across the criteria
used to identify the fiscal consolidations, namely between FC1 and the remainder. This is
actually the only criterion that has a success rate that is similar to the one found in the SU 1
case, while FC 2 and FC 3 are below their peers by more than 20 percentage points.
On the one hand, the main explanation for the difference between the success rates
within SU 2 has to do with the duration of the fiscal consolidations coupled with its lower
flexibility vis-à-vis SU 1 . As seen earlier on table I, the fiscal consolidations based on FC1
have a much higher duration than either FC 2 or FC 3 . The only requirement for a fiscal
consolidation to be successful according to SU 2 is that the level of debt-to-GDP in two years
after the end of the adjustment has to be lower than the one in the last year of the adjustment.
Therefore, longer periods of adjustment will necessarily result into more successful years of
fiscal consolidation. The same does not occur in SU 1 as it allows for successful and nonsuccessful years within the same adjustment period.19
19
This is a consequence that derives from the fact that Alesina & Ardagna (2013) treat multi-year periods as a
single episode and define all those years as either being successful or not altogether, which might have some
implications on the results. Perotti (2012) provides a detailed description of this issue.
21
On the other hand, since the success rates based on SU 1 are generally higher than in
SU 2 , one can argue that, based on these results, countries have been more successful in
improving their fiscal position rather than their levels of debt ratios. This signals that although
there are improvements in the CAPB during the fiscal consolidation periods, this doesn’t
necessarily result in a fiscal surplus at least in the next two years after the end of the
adjustment. This ultimately impinges on the countries’ debt ratios during that period.
In table VII we present some facts about the expenditure and revenue based
consolidations, which were computed for the three different criteria used to identify the fiscal
contractions. By disentangling those we can assess the possible differences regarding the
criteria used to define fiscal consolidations as successful and also the possible implications for
GDP growth. This table shows the results computed for a threshold of  =2/3, while the
results for the other thresholds can be consulted in tables XXI and XXII in the Appendix.
Table VII – Expenditure and revenue based consolidations:  =2/3
Total
Events
Expenditure
Based 1
Expenditure
Based 2
Expenditure
Based 3
Revenue
Based 1
Revenue
Based 2
Revenue
Based 3
Average Size
of the
Consolidation
1
 b
i 0
t i
Debtt  2  Debtt
23
2.36
2.87
-2.50
23
2.83
3.07
-0.74
18
3.49
3.65
-0.34
13
0.84
1.91
-1.06
15
1.64
1.32
4.49
13
1.93
1.71
4.77
Source: author’s computations.
We can see that in our panel there are significantly more expenditure based
consolidation events, rather than revenue based ones. The number of expenditure based
consolidations ranges between 18, in FC 3 and 23, both in criteria one and two. The revenue
based consolidations account for 13 and 15 events, based on both FC1 and FC 3 , and FC 2
respectively.
The average size of the consolidations (based on the change in the CAPB) is also
higher in the expenditure based cases across all the different criteria, compared to revenue
based ones. The minimum difference is 1.19 percentage points (pp.) in case 2, but it can get as
22
high as 1.56 pp. in case 3. Overall we have stronger adjustments for the expenditure based
consolidations. These findings differ from Afonso & Jalles (2012), which report no significant
difference between the size of the consolidation that is done via the expenditure or the
revenue side of the budget.
1
The next column (  bt i ) reports the changes in the cyclically adjusted primary
i 0
balance for two consecutive years, used in SU 1 . The differences between the expenditure and
revenue based consolidations lie roughly between 0.96 and 1.94 pp. in cases one and three,
respectively.
We can also look at the difference between the debt-to-GDP ratio two years after the
end of the adjustment and in the last year of the adjustment, in the following column, which
was used to compute SU 2 . All of the criteria report that on average the expenditure based
consolidations led to a decrease in the debt-to-GDP ratio on that period. On the revenue based
side, we have significant differences between FC1 and the remainder. While in the latter there
was a decrease in the debt-to-GDP ratio, in FC 2 and FC 3 it actually increased more than 4
percentage points. Nevertheless, even in the first case we can see that there was a more
significant improvement in the debt-to-GDP ratio for the expenditure based consolidation.20
Following the results in table VII, we estimated a probit model based on Afonso &
Jalles (2012) in order to assess if the reported differences between the expenditure and
revenue based consolidations are statistically relevant and impinge on the success of the fiscal
adjustments:
Pri  SU  1| Zi   E[SU  1| Zi ]    Zi 
(7)
where E[SU  1| Zi ] is the conditional expectation of the success of the fiscal consolidation,
given Z i and SU refers to the dummy variables defined in (4) and (5). Z i is defined as follows:
Zi  1   2 Di   3bi   4 EXPi  5 MX i
(8)
This is the only reported finding that doesn’t hold across the 3 different thresholds, being true only for  =2/3
and  =3/4. For  =1/2 there is no significant difference between the expenditure and revenue based episodes in
20
FC1 on this matter.
23
Di is the duration of the fiscal consolidation, bi refers to the change in the cyclically adjusted
primary balance, which accounts for the size of the consolidation. EXPi was defined in (6) as
a dummy variable that accounts for expenditure based consolidations, according to different
thresholds, while the same was done on the revenue side.
We also included MX i which refers to the dummy variable used to identify the
monetary expansions computed earlier, according to (1). The motivation behind this addition
has to do with an issue raised in the recent literature, which has to do with the possible
influence of the monetary expansions in determining the success of fiscal consolidations.
For instance, Devries et al. (2011) suggest that expenditure based consolidations
where more successful because they were complemented by monetary expansions, in the form
of strong currency devaluations. Alesina et al. (2012) mention the importance of
accompanying monetary policy in determining the possible heterogeneous effects of
expenditure based and revenue based consolidations. Alesina & Ardagna (2013) also account
for the possible role of the monetary policy in differentiating the effects of expenditure versus
revenue based adjustments.21
Table VIII shows the results for the success measure constructed by Afonso & Jalles
(2012), based on FC 2 .22 The results for the other criteria used to compute fiscal
consolidations can be consulted in tables XXIII and XXVI in the Appendix.
We can see that according to the measure first computed by Afonso & Jalles (2012)
the success of the fiscal consolidations seems to be enhanced if based on expenditure cuts. On
the other hand, we find no statistically significant results for the revenue based consolidations.
Moreover, both the duration and size of the consolidations seem to play a significant role:
longer and stronger consolidations appear to contribute positively for the success of the fiscal
consolidations. These results hold for all of the reported thresholds in FC 2 and FC 3 . In the
FC1 case we find statistically significant results only for the size of the consolidations.23
Concerning the role of the monetary policy, we find no statistically significant results.24
21
22
However, in this case they found no significant impact of monetary policy.
Some observations were excluded due to the fact that they occur in the last years of the sample and therefore
we cannot assess if they were successful according to either (4) or (5).
23
See tables XXIII and XXVI in the Appendix for FC1 and FC 3 .
Results for MX 3 are available on request and do not alter the overall findings. We could not compute the
estimations for MX 1 since it perfectly predicts the success of the fiscal consolidations.
24
24
Table VIII – Success of fiscal consolidations for SU 1 based on FC 2
Expenditure
Specification
constant
duration
Δcapb
exp12
1
-4.930***
(-3.17)
1.177**
(1.99)
1.178***
(3.34)
1.443***
(3.14)
exp23
2
-3.842***
(-2.93)
0.974*
(1.93)
1.006***
(3.23)
Revenue
3
-3.851***
(-2.97)
0.975*
(1.94)
1.009***
(3.29)
4
-3.171***
(-2.79)
0.860*
(1.87)
0.873***
(3.01)
5
-2.962**
(-2.48)
0.828*
(1.72)
0.870***
(2.89)
0.783*
(1.70)
exp34
0.783*
(1.70)
rev12
0.059
(0.12)
rev23
-0.296
(-0.62)
rev34
mx2
R2
N
6
-2.962**
(-2.48)
0.828*
(1.72)
0.870***
(2.89)
0.031
(0.06)
0.487
50
0.055
(0.11)
0.414
50
0.059
(0.11)
0.414
50
0.343
(0.57)
0.381
50
0.243
(0.46)
0.386
50
-0.296
(-0.62)
0.243
(0.46)
0.386
50
Notes: Used robust heteroskedastic-consistent standard errors. The t-statistics are in parentheses. *, ** and
*** denotes statistically significant at a 10, 5 and 1 percent level, respectively. 12, 23 and 34 next to exp and
rev refer to the relevant value for  , according to (6).
Table IX shows the results for the success criterion SU 2 , based on FC1 . The results
for FC 2 and FC 3 can be consulted in tables XXV and XXVI the Appendix. The results are
similar to the ones found in the SU 1 case, concerning the role of the duration and of the
expenditure based adjustments in the success of the fiscal consolidations. Moreover, we have
found some evidence that the revenue based consolidations have a negative impact on the
success of the adjustment. On the other hand, contrary to the findings for SU 1 , it seems that
the size of the consolidation has a negative impact on the success of the consolidation, thus
not being robust across the different criteria.
25
Table IX - Success of fiscal consolidations for SU 2 based on FC 1
Expenditure
Specification
constant
duration
Δcapb
exp12
1
-1.672
(-1.41)
0.821**
(2.56)
-0.262**
(-2.03)
1.600**
(2.48)
exp23
2
-1.690
(-1.42)
0.832***
(2.61)
-0.263**
(-2.04)
Revenue
3
-1.884
(-1.41)
0.860**
(2.42)
-0.221*
(-1.87)
4
-0.850
(-0.90)
0.720***
(3.06)
-0.233*
(-1.70)
5
-0.519
(-0.52)
0.923***
(3.41)
-0.404***
(-2.65)
1.601**
(2.46)
exp34
1.938**
(2.49)
rev12
-0.317
(-0.55)
rev23
-1.571**
(-1.99)
rev34
mx2
N
6
-0.519
(-0.52)
0.923***
(3.41)
-0.404***
(-2.65)
-0.382
(-0.69)
0.478
39
-0.389
(-0.70)
0.477
39
-0.238
(-0.40)
0.511
39
-0.175
(-0.33)
0.352
39
-0.410
(-1.67)
0.440
39
-1.571**
(-1.99)
-0.410
(-1.67)
0.440
39
Notes: Used robust heteroskedastic-consistent standard errors. The t-statistics are in parentheses. *, ** and
*** denotes statistically significant at a 10, 5 and 1 percent level, respectively. 12, 23 and 34 next to exp and
rev refer to the relevant value for  , according to (6).
Regarding the role of the monetary policy if we look at the FC 2 case in table XXV in
the Appendix, there seems to be some evidence that real currency devaluations ( MX 2 )
contribute negatively to the success of the adjustments. However, since we cannot check the
robustness of these results with a monetary expansion based on the real short term interest
rate ( MX 1 ) because of the same problem reported earlier for SU 1 , we would not extract a
clear conclusion here. Furthermore, the fact that MX 1 perfectly predicts the success of the
fiscal consolidations could actually lead to opposite conclusions to the ones found for either
MX 2 or MX 3 . So we would rather say that the impact of the monetary easing in the success
of the fiscal consolidations is not clear.
To sum up, the most robust findings for the success of the fiscal consolidation were
obtained for the impact of the duration and of the expenditure based consolidations. Both
contribute positively for the success of the fiscal adjustments, across the different criteria. In
addition, there is some evidence that fiscal consolidations based on tax raises have a negative
impact on the success of the fiscal consolidations.
26
The size of the consolidation gives us mixed evidence: it seems to contribute
positively for the success of fiscal consolidations based on SU 1 , which is consistent with
Afonso & Jalles (2012), but the opposite is verified for SU 2 . The role of the monetary policy
is also unclear.
Table XXVII in the appendix shows the summary of the robustness tests for the
estimations computed for specification (8).
5. Conclusions
This paper aimed at providing new insights about expansionary fiscal consolidations
in the EMU, by incorporating monetary developments on specifications previously used on
empirical research. The Fixed Effects panel estimations conducted for 14 European Union
countries show no evidence of non-Keynesian effects during fiscal consolidations, when the
monetary policy developments are not considered. Nevertheless, there is some evidence of
non-Keynesian effects in the absence of fiscal consolidations.
On the other hand, when the baseline specification is extended in order to
accommodate the monetary developments, there is some evidence of non-Keynesian effects
during fiscal consolidations. When fiscal consolidation episodes are matched by a monetary
expansion, there is a shift on the standard Keynesian impact of government final consumption
expenditure and taxation on private consumption.
Overall, when fiscal consolidations are not matched by a monetary expansion, the nonKeynesian effects captured earlier disappear. The size of the increase in private consumption
due to a fiscal consolidation depends on the absence of liquidity-constrained households,
which may prevent Ricardian behaviour, thus undermining the permanent income hypothesis
of consumption smoothing. A monetary expansion provides the necessary liquidity increase
during fiscal consolidations that allows individuals to smooth their consumption.
Attending on the success of the fiscal consolidations, countries have been more
effective in improving their fiscal position rather than their levels of debt ratios.
Improvements in the CAPB during the fiscal consolidation periods do not necessarily result in
a fiscal surplus, at least in the next two years after the end of the adjustment, which ultimately
impinges on the countries’ debt ratios.
Generally, we have stronger adjustments for the expenditure based consolidations as
their size is significantly higher vis à vis revenue based ones.
27
The probit estimations show evidence which suggests that longer lasting adjustment
periods seem to contribute positively for their success. Even so, the role of the size of the
consolidations in this regard is unclear.
Additionally, expenditure based consolidations are more likely to be successful than
the ones based on tax raises. These findings are more robust for the expenditure based
consolidations.
The overall role of the monetary policy in the success of the fiscal consolidations is
unclear. On the one hand, we have some (although scarce) evidence that monetary expansions
based on real currency devaluations contribute negatively for the success of fiscal
consolidations. On the other hand, we cannot perform probit estimations for monetary
expansions based on the real interest rate since those near to perfectly predict the success of
fiscal consolidations, which means that in almost every case a monetary expansion based on
the real interest rate is associated with a successful fiscal adjustment.
Future research may include the assessment of the factors that may influence the
occurrence of expansionary episodes, through a binary choice model and also the usage of the
so called policy-action based approach for identifying fiscal episodes.
28
Appendix
Table X – Data sources
Original Series
Total population, thousands.
Gross domestic product, millions, national currency, current market prices.
Price deflator of gross domestic product, national currency, 2005=100.
Private final consumption expenditure at 2005 constant prices, millions, national currency.
Final consumption expenditure of general government at 2005 constant prices, millions, national
currency.
Social benefits other than social transfers in kind, general government, millions, national
currency, current prices.
Current taxes on income and wealth (direct taxes), general government, millions, national
currency, current prices.
Total expenditure: general government: ESA 1995 (Including one-off proceeds (treated as
negative expenditure) relative to the allocation of mobile phone licenses (UMTS)).
Total revenue: general government: ESA 1995.
General government consolidated gross debt: Excessive deficit procedure (based on ESA 1995)
and former definition (linked series); % GDP
Taxes linked to imports and production (indirect taxes), general government, millions, national
currency, current prices.
Net borrowing (+) or net lending (-) excluding interest of general government adjusted for the
cyclical component. Adjustment based on potential GDP excessive deficit procedure (% of GDP
at market prices).
Real short-term interest rates, deflator private consumption.
Nominal Effective exchange rate 2005=100: Performance relative to the rest of 24 industrial
countries: double export weights: EU-15, TR CH NR US CA JP AU MX and NZ.
Real effective exchange rate, consumer price index deflated; 2005=100; IMF Statistics Database
AMECO
Code
NPTN
UVGD
PVGD
OCPH
OCTG
UYTGH
UTYG
UUTG
URTG
UDGGL
UTVG
UBLGBP
ISRC
XUNNQ
Table XI – Descriptive statistics of the variables used to identify the fiscal and monetary
episodes
Variable
Mean
Maximum
Minimum
Std. Dev.
N
bt
-0.01
16.41
-15.62
2.01
454
M
1
-0.03
10.61
-15.66
2.43
558
M
2
0.18
26.62
-17.92
4.42
501
M
3
-0.52
16.73
-21.40
4.88
588
Source: authors’ computations. Notes: bt – Change in cyclically adjusted
primary balance; M 1 – Absolute change of the real short term interest rate;
M 2 – Percent change of the real effective exchange rate; M 3 – Percent
change of the nominal effective exchange rate; All indicators were computed
based on annual data.
29
Table XII – Unit root tests for the series used in the fixed effects estimations
C
C
Y
Y
Y av
Y av
FCE
FCE
TF
TF
TAX
TAX
M 1
M 2
M 3
Common Unit Root (LLC)
t-stat.
p-value
N
-7.14
0.00
574
-3.31
0.00
560
-6.01
0.00
574
-9.83
0.00
560
-6.74
0.00
574
-12.41
0.00
560
-9.43
0.00
574
-4.28
0.00
560
-7.42
0.00
456
-7.89
0.00
442
-5.02
0.00
456
-8.83
0.00
442
-12.47
0.00
530
-9.24
0.00
473
-9.66
0.00
560
Individual Unit Root (IPS)
t-stat.
p-value
N
-2.02
0.02
574
-6.70
0.00
560
-1.11
0.13
574
-9.74
0.00
560
-1.20
0.12
574
-9.01
0.00
560
-4.86
0.00
574
-5.01
0.00
560
-2.32
0.01
456
-7.72
0.00
442
-0.95
0.17
456
-8.37
0.00
442
-15.70
0.00
530
-9.29
0.00
473
-10.73
0.00
560
Source: authors’ computations. Notes: LLC – Levin, Lin and Choo test;
IPS – Im, Pesaran and Chin test.
Table XIII - Descriptive statistics for the series used in the probit estimations
Series
Mean
Median Maximum Minimum
Std. Dev.
Observations
bt
-0.01
-0.04
16.41
-15.62
2.00
456
Exp
0.22
0.07
17.35
-18.41
2.25
470
Rev
0.17
0.24
4.44
-4.24
1.21
470
Debt
57.45
55.02
170.31
6.15
29.15
586
Source: authors’ computations. Notes: bt – Change in the cyclically adjusted
primary balance; Exp – Change in the general government expenditure; Rev
– Change in general government revenue; Debt – General government gross
debt; All variables are expressed as GDP ratios.
30
Table XIV – Fixed Effects estimation for specification (3): 2nd output

Ct 1
0
Yt 1
1
Yt
0
Yt av
1
1
Yt av
10
FCEt 1
 30
FCEt
10
TFt 1
 30
TFt
 10
TAX t 1
 30
TAX t
50
M tl
 FC m
 MX l
FC 1 , MX 1
FC 1 , MX 2
FC 2 , MX 2
-0.097***
(-4.05)
0.103***
(3.77)
0.795***
(11.01)
-0.030**
(-2.13)
-0.148**
(-2.14)
0.984***
(6.73)
-2.288***
(-6.53)
-0.275***
(-5.74)
-3.041***
(-6.35)
-0.621***
(-6.50)
-1.98***
(-5.94)
0.024***
(5.59)
-0.091***
(-3.76)
0.079***
(3.00)
0.740***
(10.89)
-0.024
(-1.61)
-0.136**
(-1.97)
0.069*
(1.69)
-0.161
(-0.89)
0.007
(0.50)
0.089
(0.44)
-0.056
(-1.46)
0.057
(0.41)
0.098
(0.83)
-0.090***
(-3.74)
0.077***
(2.95)
0.745***
(10.83)
-0.024
(-1.60)
-0.133*
(-1.92)
0.107
(1.91)
-0.124
(-0.70)
0.012
(0.822)
-0.022
(-0.09)
-0.091*
(-1.68)
0.049
(0.33)
0.076
(0.57)
Notes: Used robust heteroskedastic-consistent standard errors. The t-statistics are in parentheses. *, ** and
*** denotes statistically significant at a 10, 5 and 1 percent level, respectively.
31
Table XV – Fixed Effects estimation for specification (3): 2nd output (cont.)
 20
FCEt 1
 40
FCEt
 20
TFt 1
 (1  FC m )
FC 1 , MX 1
-0.036**
(-2.19)
0.023
(0.20)
-0.019
(-1.62)
-0.041
(-0.85)
0.055***
(3.64)
FC 2 , MX 2
-0.007
(-0.34)
-0.062
(-1.02)
-0.022**
(-2.24)
-0.083*
(-1.73)
0.041***
(2.93)
FC 2 , MX 2
-0.003
(-0.13)
-0.060
(-1.03)
-0.022**
(-2.36)
-0.073
(-1.58)
0.039***
(2.87)
 40
TFt
 20
TAX t 1
 40
TAX t
0.013
(0.28)
0.013
(0.33)
0.015
(0.41)
60
M tl
-0.001
(-0.92)
0.098
(2.41)
0.107***
(2.64)
11
FCEt 1
-0.006
(-0.23)
-0.021
(-0.72)
-0.006
(-0.29)
 31
FCEt
0.321***
(3.54)
0.363***
(4.16)
0.375***
(4.80)
11
TFt 1
 FC m
0.011
(0.71)
0.005
(0.32)
 31
TFt
 (1  MX l )
0.007
(0.50)
-0.076
(-0.97)
-0.110
(-1.24)
-0.100
(-0.99)
 11
TAX t 1
0.006
(0.26)
0.025
(1.11)
0.019
(0.98)
 31
TAX t
0.027
(0.48)
0.140**
(2.47)
0.067
(1.60)
51
M tl
0.002
(1.32)
0.028
(0.49)
0.032
(0.55)
 21
FCEt 1
-0.014
(-0.94)
-0.015
(-1.03)
-0.016
(-1.14)
 41
FCEt
0.039
(0.58)
0.091
(1.29)
0.083
(1.14)
 21
TFt 1
0.005
(0.74)
0.008
(1.15)
0.008
(1.24)
 41
TFt
0.060*
(1.69)
0.046
(1.33)
0.050
(1.49)
 21
TAX t 1
0.015
(1.44)
0.024**
(2.20)
0.026**
(2.36)
 41
TAX t
0.043*
(1.78)
0.056**
(2.46)
0.054**
(2.28)
61
M tl
-0.001
(-1.01)
454
0.758
t-stat.
p-val.
0.037
(1.39)
454
0.755
t-stat.
p-val.
0.040
(1.49)
454
0.779
t-stat.
p-val.
4.03
0.00
3.62
0.00
3.46
0.00
N
Redundant FE
likelihood ratio
 MX l
 (1  FC m )
 (1  MX l )
Notes: Used robust heteroskedastic-consistent standard errors. The t-statistics are in parentheses. *, ** and
*** denotes statistically significant at a 10, 5 and 1 percent level, respectively.
32
Table XVI – Fixed Effects estimation for specification (3): 3rd output

Ct 1
0
Yt 1
1
Yt
0
Yt av
1
1
Yt av
10
FCEt 1
 30
FCEt
10
TFt 1
 30
TFt
 10
TAX t 1
 30
TAX t
50
M tl
 FC m
 MX l
FC 2 , MX 3
FC 3 , MX 2
FC 3 , MX 3
-0.093***
(-4.05)
0.091***
(3.15)
0.810***
(11.05)
-0.016
(-0.96)
-0.166**
(-2.29)
0.031
(0.83)
0.000
(0.00)
-0.020
(-1.16)
0.038
(0.45)
-0.015
(-0.56)
-0.010
(-0.11)
0.172
(2.47)
-0.084***
(-3.51)
0.075***
(2.84)
0.766***
(11.30)
-0.026*
(-1.72)
-0.151**
(-2.17)
0.278**
(2.04)
-0.576
(-1.27)
-0.008
(-0.31)
-0.873
(-0.98)
-0.229*
(-1.79)
-0.338
(-1.11)
-0.095
(-0.39)
-0.092***
(-3.63)
0.093***
(3.26)
0.804***
(10.87)
-0.017
(-1.06)
-0.167**
(-2.30)
0.052
(1.35)
0.084
(0.59)
-0.033
(-1.57)
-0.079
(-0.51)
-0.019
(-0.62)
-0.103
(-0.84)
0.220***
(2.79)
Notes: Used robust heteroskedastic-consistent standard errors. The t-statistics are in parentheses. *, ** and
*** denotes statistically significant at a 10, 5 and 1 percent level, respectively.
33
Table XVII – Fixed Effects estimation for specification (3): 3rd output (cont.)
 20
FCEt 1
 40
FCEt
 20
TFt 1
 (1  FC m )
FC 2 , MX 3
-0.008
(-0.37)
0.193*
(1.83)
0.011
(1.05)
-0.034
(-0.80)
-0.007
(-0.41)
FC 3 , MX 2
-0.003
(-0.16)
-0.062
(-1.03)
-0.018*
(-1.89)
-0.046
(-0.93)
0.035**
(2.54)
FC 3 , MX 3
-0.006
(-0.28)
0.167*
(1.68)
0.008
(0.82)
-0.029
(-0.82)
-0.007
(-0.44)
 40
TFt
 20
TAX t 1
 40
TAX t
-0.038
(-0.67)
0.030
(0.75)
-0.048
(-0.85)
60
M tl
0.035
(0.83)
0.090**
(2.26)
0.048
(1.18)
11
FCEt 1
0.001
(0.04)
-0.001
(-0.03)
0.000
(0.02)
 31
FCEt
0.254**
(2.14)
0.375***
(5.07)
0.318**
(2.58)
11
TFt 1
 FC m
-0.008
(-0.51)
-0.003
(-0.14)
 31
TFt
 (1  MX l )
0.004
(0.22)
-0.041
(-0.41)
-0.106
(-1.01)
-0.051
(-0.48)
 11
TAX t 1
-0.009
(-0.41)
0.024
(1.16)
-0.004
(-0.20)
 31
TAX t
0.066
(1.16)
0.053
(1.16)
0.090
(1.40)
51
M tl
0.065
(1.08)
-0.010
(-0.12)
0.022
(0.21)
 21
FCEt 1
-0.024*
(-1.67)
-0.016
(-1.12)
-0.024*
(-1.72)
 41
FCEt
0.003
(0.05)
0.079
(1.10)
0.003
(0.05)
 21
TFt 1
0.002
(0.26)
0.008
(1.14)
0.002
(0.31)
 41
TFt
0.042
(1.14)
0.049
(1.50)
0.039
(1.06)
 21
TAX t 1
0.017
(1.29)
0.026**
(2.31)
0.016
(1.23)
 41
TAX t
0.024
(0.94)
0.054**
(2.28)
0.024
(0.97)
61
M tl
0.035
(1.21)
454
0.752
t-stat.
p-val.
0.043
(1.78)
454
0.780
t-stat.
p-val.
0.038
(1.47)
454
0.756
t-stat.
p-val.
3.26
0.00
3.33
0.00
3.29
0.00
N
Redundant FE
likelihood ratio
 MX l
 (1  FC m )
 (1  MX l )
Notes: Used robust heteroskedastic-consistent standard errors. The t-statistics are in parentheses. *, ** and
*** denotes statistically significant at a 10, 5 and 1 percent level, respectively.
34
Table XVIII – Wald coefficient diagnostics for estimations based on specification (3)
Null Hypothesis
10  11  0
30  31  0
10  11  0
30  31  0
 10   11  0
 30   31  0
FC 1 , MX 3
t-stat.
1.03
-3.93
0.86
-0.32
-1.53
-2.95
1
Null Hypothesis
10  11  0
30  31  0
10  11  0
30  31  0
 10   11  0
 30   31  0
FC , MX
t-stat.
6.43
-7.14
-5.49
-6.05
-6.67
-5.93
2
Null Hypothesis
10  11  0
30  31  0
10  11  0
30  31  0
 10   11  0
 30   31  0
FC , MX
t-stat.
0.81
-1.45
-1.08
0.60
-0.21
-0.74
p-val.
0.30
0.00
0.39
0.75
0.13
0.00
FC 2 , MX 1
t-stat.
1.28
-2.97
0.69
0.20
-1.73
4.38
1
1
p-val.
0.00
0.00
0.00
0.00
0.00
0.00
FC , MX
t-stat.
1.98
-2.60
-0.21
0.91
-1.94
-0.55
3
3
p-val.
0.42
0.15
0.28
0.55
0.83
0.46
FC , MX
t-stat.
2.02
-2.06
0.00
-0.85
-1.99
-1.27
p-val.
0.20
0.00
0.49
0.84
0.08
0.00
FC 3 , MX 1
t-stat.
-14.01
-3.02
21.64
-19.08
-9.75
16.32
2
2
p-val.
0.05
0.01
0.83
0.36
0.05
0.58
FC , MX
t-stat.
1.96
-2.58
0.36
0.30
-2.00
-0.12
3
2
p-val.
0.04
0.04
1.00
0.39
0.05
0.20
FC , MX
t-stat.
1.30
-1.31
-1.15
-0.15
-0.40
-1.40
p-val.
0.00
0.00
0.00
0.00
0.00
0.00
2
p-val.
0.05
0.01
0.72
0.76
0.05
0.91
3
p-val.
0.19
0.19
0.25
0.88
0.69
0.16
Notes: Wald coefficient diagnostics for the estimations on tables IV-V, XIV-XV and XVI-XVII respectively.
35
Table XIX – Robustness tests for estimations based on specification (3)
Sample restriction
Summary results
Sample with
"central-European"
countries
Some evidence of non-keynesian effects for taxes during fiscal consolidations,
which holds both in the presence and absence of monetary expansions. Evidence
seems stronger when fiscal consolidations are matched by monetary expansions.
Could not compute many estimations due to near singular matrix problems.
Sample with "peripheral"
countries
Some evidence of non-keynesian effects for all of the budgetary components,
which seems to be stronger when fiscal consolidations are matched by monetary
expansions. Could not compute some estimations due to near singular matrix
problems.
1970 – 1998
Some evidence of non-keynesian effects for all of the budgetary components,
which seems to be stronger when fiscal consolidations are matched by monetary
expansions.
1999 – 2012
We could not compute any estimation due to near singular matrix.
Notes: “Central-European” countries include all but Greece, Ireland, Italy, Portugal and Spain, which are
labelled as peripheral countries. Estimations are available on request.
36
Table XX – Successful fiscal consolidations according to the different criteria
(1970-2012)
FC 1
SU 1
FC 2
84, 05
82-84
83-86
88, 96-97,
00
FC 3
84, 05
82, 84
83-86
88, 96, 00
82-87
83-87
97-98
FC 1
SU 2
FC 2
01, 05
FC 3
01, 05
83-86
88, 96-97
83-86
88, 96, 00
88
88
93, 96
93, 96
86, 88
86, 88
Country
Austria
Belgium
Denmark
Finland
82-84
83-86
97, 00
France
Germany
96, 99
96, 11
96, 11
96-99
91, 94, 1011
88, 11
82, 92
91, 94,
10-11
88, 11
82, 92
96
Ireland
Italy
93-94,
10-11
11
92
92-94
Netherlands
91
91
91
93
Portugal
83, 11
83, 88, 11
83, 88, 11
97
97-99, 11
96-97
97-98, 11
96-97
11
97-99
97-00
96-97
97-98, 00
96-97
00
25
32
28
29
19
17
Greece
Spain
Sweden
United
Kingdom
#
Successful
years
Source: author’s computations. Notes: SU 1 - Success measure based on Afonso & Jalles (2012); SU 2 Success measure based on Alesina & Ardagna (2013); FC 1 - Measure based on Giavazzi & Pagano
(1996); FC 2 - Measure based on Alesina & Ardagna (1998); FC 3 - Measure based on Afonso (2006,
2010).
Table XXI – Expenditure and revenue based consolidations:  =1/2
Average Size
of the
Consolidation
 b
24
2.34
2.94
-2.50
27
2.75
3.07
-0.85
22
3.25
3.53
-1.08
19
1.28
2.24
-2.62
22
1.94
2.13
2.07
19
2.17
2.39
2.61
Total
Events
Expenditure
Based 1
Expenditure
Based 2
Expenditure
Based 3
Revenue
Based 1
Revenue
Based 2
Revenue
Based 3
Source: authors’ computations.
37
1
i 0
t i
Debtt  2  Debtt
Table XXII – Expenditure and revenue based consolidations:  =3/4
Average Size
of the
Consolidation
 b
21
2.34
2.86
-2.52
22
2.75
3.07
-0.64
17
3.42
3.68
-0.16
12
0.71
1.91
-1.06
14
1.58
1.32
4.49
12
1.89
1.71
4.77
Total
Events
Expenditure
Based 1
Expenditure
Based 2
Expenditure
Based 3
Revenue
Based 1
Revenue
Based 2
Revenue
Based 3
1
t i
i 0
Debtt  2  Debtt
Source: authors’ computations.
Table XXIII – Success of fiscal consolidations for SU 1 based on FC 1
Expenditure
Specification
constant
duration
Δcapb
exp12
1
-1.229*
(-1.79)
0.184
(1.12)
0.516***
(3.85)
-0.080
(-0.18)
exp23
2
-1.182*
(-1.73)
0.186
(1.14)
0.511***
(3.87)
Revenue
3
-1.091
(-1.61)
0.197
(1.18)
0.500***
(3.83)
4
-1.601**
(-2.14)
0.208
(1.31)
0.548***
(3.85)
5
-1.660**
(-2.16)
0.218
(1.34)
0.579***
(3.58)
-0.179
(-0.40)
exp34
-0.433
(-0.95)
rev12
0.416
(0.91)
rev23
0.544
(1.00)
rev34
mx2
N
6
-1.660**
(-2.16)
0.218
(1.34)
0.579***
(3.58)
-0.188
(-0.38)
0.303
43
-0.175
(-0.35)
0.305
43
-0.150
(-0.30)
0.317
43
-0.111
(-0.22)
0.315
43
-0.111
(-0.21)
0.320
43
0.544
(1.00)
-0.111
(-0.21)
0.320
43
Notes: Used robust heteroskedastic-consistent standard errors. The t-statistics are in parentheses. *, ** and
*** denotes statistically significant at a 10, 5 and 1 percent level, respectively. 12, 23 and 34 next to exp and
rev refer to the relevant value for  , according to (6).
38
Table XXIV – Success of fiscal consolidations for SU 1 based on FC 3
Expenditure
Specification
constant
1
-5.197***
(-2.74)
duration
Δcapb
exp12
2.052***
(2.73)
1.362***
(2.70)
exp23
2
-14.678***
(-2.61)
3.375**
(2.02)
4.596***
(2.73)
Revenue
3
-14.678***
(-2.61)
3.375**
(2.02)
4.596***
(2.73)
4
-10.752***
(-2.84)
2.447**
(2.19)
3.688***
(2.89)
5
-10.622***
(-3.11)
2.481**
(2.35)
3.586***
(3.21)
6
-10.622***
(-3.11)
2.481**
(2.35)
3.586***
(3.21)
1.375*
(1.74)
exp34
1.375*
(1.74)
rev12
-0.704
(-1.19)
rev23
-0.852
(-1.23)
rev34
mx2
-0.180
(-0.28)
0.522
-0.373
(-0.49)
0.603
-0.373
(-0.49)
0.603
-0.024
(-0.03)
0.566
0.060
(0.09)
0.577
-0.852
(-1.23)
0.060
(0.09)
0.577
N
42
42
42
42
42
42
Notes: Used robust heteroskedastic-consistent standard errors. The t-statistics are in parentheses. *, ** and
*** denotes statistically significant at a 10, 5 and 1 percent level, respectively. 12, 23 and 34 next to exp and
rev refer to the relevant value for  , according to (6).
39
Table XXV – Success of fiscal consolidations for SU 2 based on FC 2
Expenditure
Specification
constant
duration
Δcapb
exp12
1
-0.293
(-0.51)
0.011
(0.06)
-0.040
(-0.33)
0.860**
(2.12)
exp23
2
0.019
(0.04)
0.003
(0.02)
-0.054
(-0.46)
Revenue
3
-0.011
(-0.02)
0.003
(0.02)
-0.048
(-0.40)
4
0.249
(0.42)
-0.008
(-0.04)
-0.067
(-0.55)
5
0.931
(1.41)
-0.124
(-0.64)
-0.130
(-1.03)
0.356
(0.88)
exp34
0.397
(1.64)
rev12
-0.115
(-0.271)
rev23
-1.121**
(-2.15)
rev34
mx2
N
6
0.931
(1.41)
-0.124
(-0.64)
-0.130
(-1.03)
-1.215**
(-2.40)
0.136
45
-1.093**
(-2.20)
0.08
45
-1.078**
(-2.22)
0.08
45
-1.011*
(-1.89)
0.07
45
-1.308**
(-2.44)
0.146
45
-1.121**
(-2.15)
-1.308**
(-2.44)
0.146
45
Notes: Used robust heteroskedastic-consistent standard errors. The t-statistics are in parentheses. *, ** and
*** denotes statistically significant at a 10, 5 and 1 percent level, respectively. 12, 23 and 34 next to exp and
rev refer to the relevant value for  , according to (6).
40
Table XXVI – Success of fiscal consolidations for SU 2 based on FC 3
Expenditure
Specification
constant
duration
Δcapb
exp12
1
-0.477
(-0.71)
0.176
(0.78)
-0.097
(-0.62)
1.036**
(2.37)
exp23
2
-0.076
(-0.12)
0.147
(0.70)
-0.116
(-0.78)
Revenue
3
-0.004
(-0.01)
0.137
(0.66)
-0.104
(-0.69)
4
0.263
(0.38)
0.149
(0.71)
-0.140
(-0.87)
5
0.815
(1.18)
0.050
(0.24)
-0.202
(-1.29)
0.482
(1.10)
exp34
0.187
(0.42)
rev12
-0.246
(-0.53)
rev23
-0.993*
(-1.94)
rev34
mx2
N
6
0.815
(1.18)
0.050
(0.24)
-0.202
(-1.29)
-0.765
(-1.30)
0.142
37
-0.679
(-1.19)
0.057
37
-0.534
(-0.90)
0.039
37
-0.602
(-0.98)
0.041
37
-0.786
(-1.28)
0.109
37
-0.993*
(-1.94)
-0.786
(-1.28)
0.109
37
Notes: Used robust heteroskedastic-consistent standard errors. The t-statistics are in parentheses. *, ** and
*** denotes statistically significant at a 10, 5 and 1 percent level, respectively. 12, 23 and 34 next to exp and
rev refer to the relevant value for  , according to (6).
41
Table XXVII – Robustness tests for estimations based on specification (8)
Sample restriction
Summary results
Sample with
"central-European"
countries
Similar pattern, compared to the unrestricted case.
Sample with "periphery"
countries
We could not compute any estimation, except for the SU2 case, due to lack of
variability of some variables. Didn't get statistically significant results except for
expenditure based adjustments. Those coefficients have the same signal as in the
unrestricted case.
1970 – 1998
Similar pattern for the SU1 case. In the SU2 case, the results are similar to the ones
found for the unrestricted sample, except for the size of the consolidations, which
didn't turn any statistically significant results.
1999 – 2012
We could not compute any estimation due to lack of variability of some variables.
Notes: “Central-European” countries include all but Greece, Ireland, Italy, Portugal and Spain, which are
labelled as periphery countries. Estimations are available on request.
42
References
Afonso, A. (2006). Expansionary fiscal consolidations in Europe: new evidence. European
Central Bank, Woking Paper Series No. 675, September.
Afonso, A. (2010). Expansionary fiscal consolidations in Europe: new evidence. Applied
Economics Letters 17 (2), 105-109.
Afonso, A. and Jalles, J. (2014). Assessing fiscal episodes. Economic Modelling, 37, 255-270.
Afonso, A. and Jalles, J. (2012). Measuring the success of fiscal consolidations. Applied
Financial Economics 22 (13), 1053-1061.
Afonso, A. and Sousa, R. (2011). The macroeconomic effects of fiscal policy in Portugal: a
Bayesian SVAR analysis. Portuguese Economic Journal 10 (1), 61-82.
Alesina, A. and Ardagna, S. (1998). Tales of Fiscal Adjustment. Economic Policy 27 (13),
487-545.
Alesina, A. and Ardagna, S. (2013). The Design of Fiscal Adjustments. In: Brown, J. (Ed.)
NBER book series Tax Policy and the Economy (27), 19-67.
Alesina, A. and Perotti, R. (1997). Fiscal adjustments in OECD countries: compostion and
macroeconomic effects. International Monetary Fund Staff Papers 44, 210-248.
Alesina, A., Favero, C. and Giavazzi, F. (2012). The output effects of fiscal Adjustments.
NBER Working Paper, no 18336.
Ardagna, S. (2004). Fiscal Stabilizations: When do they Work and Why. European Economic
Review 48 (5) 1047-1074.
Barro, R. (1974). Are Government Bonds Net Wealth? In: Journal of Political Economy 87,
940 - 971.
Bertola, G. and Drazen, A. (1993). Trigger Points and Budget Cuts: Explaining the Effects of
Fiscal Austerity. American Economic Review 83 (1), 11-26.
Blanchard, O. (1990). Comment, on Giavazzi and Pagano (1990). In Blanchard, O., Fischer,
S. (Eds) NBER Macroeconomics Annual (1990), 111-116.
Devries, P., Guajardo, J., Leigh, D. and Pescatori, A. (2011). An Action-based Analysis of
Fiscal Consolidations in OECD Countries. IMF Working Paper Series no. 11/128.
Giavazzi, F. and Pagano, M. (1990). Can Severe Fiscal Contractions be Expansionary? Tales
of Two Small European Countries. In Blanchard, O. and Fisher, S. (Eds.). NBER
Macroeconomics Annual (1990), MIT Press.
Giavazzi, F. and Pagano, M. (1996). Non-keynesian Effects of Fiscal Policy Changes:
International Evidence and the Swedish Experience. Swedish Economic Policy Review 3 (1),
67-103.
43
Giavazzi, F., Jappelli, T. and Pagano, M. (2000). Searching for non-linear effects of fiscal
policy: evidence from industrial and developing countries. European Economic Review 44
(7), 1259-1289.
Giudice, G., Turrini, A. and in’t Veld, J. (2004). Non-Keynesian fiscal consolidation in the
EU? Ex post evidence and ex ante analysis. CEPR Discussion Paper 4388.
Hjelm, G. (2002). Is private consumption higher (lower during periods of fiscal contractions
(expansions)? Journal of Macroeconomics 24, 17-39.
Hellwig, M. and Neumann, M. (1987). Economic Policy in Germany: Was there a
Turnaround? Economic Policy 2 (4), 103-145.
Perotti, R. (1999). Fiscal Policy in Good Times and Bad. Quarterly Journal of Economics 144
(4), 1399-1436.
Perotti R. (2012). The Austerity Myth: Growth without Pain?. In A. Alesina and F. Giavazzi
(eds.) Fiscal Policy After the Great Recession (2013) University of Chicago Press and NBER.
Prammer, D. (2004). Expansionary Fiscal Consolidations? An Appraisal of the Literature on
Non-Keynesian Effects of Fiscal Policy and a Case Study for Austria. Monetary Policy and
the Economy 3, 34-52.
Sutherland, A. (1997). Fiscal Crises and Aggregate Demand: Can High Public Debt Reverse
the Effects of Fiscal Policy? Journal of Public Economics 65 (2), 147-162.
Greene, William H. (2012). Econometric analysis, 7th ed. New Jersey: Prentice Hall.
44